def test_binary_variables_illegal_list(self):
d2 = d.copy()
for t in [[-1], [5], [0, -1], [-1, 0], [0, 10], [[-1]],
[np.array([0])]]:
d2['binary_variables'] = t
with pytest.raises(AssertionError):
Mt = SurrogateModel(d2)
def test_binary_variables_illegal_integer(self):
d2 = d.copy()
for t in [-1, 5]:
d2['binary_variables'] = t
with pytest.raises(AssertionError) as errinfo:
Mt = SurrogateModel(d2)
assert 'Value outside range' in str(errinfo.value)
def test_binary_variables_list(self):
d2 = d.copy()
d2['binary_variables'] = [1]
Mt = SurrogateModel(d2)
assert isinstance(Mt.binary_variables, list)
assert len(Mt.binary_variables) == 1
assert Mt.binary_variables[0] == 1
def test_prediction(self):
# Nothing given
Mt = SurrogateModel(d)
with pytest.raises(AssertionError):
M, S = Mt.predict(X)
# Give mean
Mt.pmean = np.random.uniform(*p_bounds.T)
with pytest.raises(AssertionError):
M, S = Mt.predict(X)
# Give training data
Mt.set_training_data(Z, Y)
with pytest.raises(AssertionError):
M, S = Mt.predict(X)
# Give hyperparameters
Mt.hyp = [0] * Mt.num_outputs
with pytest.raises(NotImplementedError):
M, S = Mt.predict(X)
def test_get_Z(self):
Mt = SurrogateModel(d)
Mt.trans_p = lambda p: p
z = (3. + np.arange(Dx + Dp))[None, :]
x = z[:, :Dx]
p = z[:, Dx:]
with pytest.raises(AssertionError):
Z = Mt.get_Z(x)
Z = Mt.get_Z(x, p)
assert np.all(Z == z)
Mt.pmean = p[0]
Z = Mt.get_Z(x)
assert np.all(Z == z)
def test_load_dict(self):
Mt = SurrogateModel(d)
Y = np.random.randn(25, Mt.num_outputs)
Mt.Y = Y
d2 = Mt._get_save_dict()
pt = np.random.randn(Mt.dim_p)
d2['pmean'] = pt.copy()
Mt._load_save_dict(d2)
assert np.all(Mt.pmean == pt)
assert np.all(np.abs(d2['Y'] - Y) <= 1e-10)
assert Mt.Z is None
def test_binary_variables_else(self):
d2 = d.copy()
for t in [0.1, None, np.array([1])]:
d2['binary_variables'] = t
with pytest.raises(ValueError):
Mt = SurrogateModel(d2)
def test_training_data(self):
self._correct_shape(M)
Mt = SurrogateModel(d)
self._is_none(Mt)
Mt.Z = Z
Mt.Y = Y
self._correct_shape(Mt)
Mt.clear_model()
self._is_none(Mt)
Mt.set_training_data(Z, Y)
self._correct_shape(Mt)
Mt.clear_model()
self._is_none(Mt)
Mt.set_training_data(X, P, Y)
self._correct_shape(Mt)
Mt.clear_model()
self._is_none(Mt)
if X.ndim == 1:
return _f(X, P)
return np.array([_f(x, p) for x, p in zip(X, P)])
Dx, Dp, E = 3, 2, 2
d = {
'name': 'testmodel',
'call': f,
'dim_x': Dx,
'dim_p': Dp,
'num_outputs': E,
'binary_variables': [2]
}
M = SurrogateModel(d)
X = np.array([[10., 5.], [10., 8.], [20., 5.], [20., 8.]])
Xs = np.random.uniform([10, 5], [20, 8], size=(25, 2))
X = np.vstack((X, Xs))
X = np.c_[X, np.random.randint(0, 1, len(X))]
P = np.array([[2., 3.], [2., 5.], [4., 3.], [4., 5.]])
Ps = np.random.uniform([2., 3.], [4., 5.], size=(len(X) - 4, 2))
P = np.vstack((P, Ps))
Z = np.c_[X, P]
Y = f(X, P)
M.set_training_data(Z, Y)
ymean = np.mean(Y, axis=0)